Changelog
What's new.
Every release, newest first, in the words members read in the app.
- Releases
- 1,155
- Days shipping
- 98
- Now on
- v5.163.11
Tuesday, Jul 28, 20261 release
Each Dose Counted Once
v4.36.1- Fixed: after editing a schedule, the Next Dose card could list the same compound twice โ the old (ended) protocol kept projecting doses alongside its replacement. Ended protocols now stop projecting at their end date, so every dose shows exactly once
- Schedule changes now fully retire the old protocol row, the same way ending a schedule manually does. Atlas, exports, and every active-schedule view see exactly one live protocol per compound โ while the blood-level chart still keeps the old protocol's full history and decay tail
Monday, Jul 27, 20263 releases
PRs Are Real Lifts, Not Math
v4.36.0- Personal records now track what you actually lifted, not an Epley projection. Weight PR = the heaviest weight you've ever put on the bar for an exercise; Rep PR = more reps than you've ever done at that weight or heavier. Put 20 more pounds on the bar and the app celebrates it โ no more 'not a PR' because a formula guessed your ceiling was higher
- PR toasts on session finish: complete a workout and every weight PR and rep PR you just hit pops up immediately โ the celebration happens at the moment it should
- Your record book keeps both kinds: heaviest lift plus your best-reps records at each meaningful weight (dominated records are pruned automatically). Session pages badge weight PRs in amber and rep PRs in blue
- Estimated 1RM didn't disappear โ it moved to where estimates belong: the strength-curve trend chart, clearly labeled as an Epley estimate. It is never again the gatekeeper of what counts as a record
- Entire PR history rebuilt from your real sets: every past workout was replayed chronologically through the new detection, so records you hit months ago (and the app never credited) now show with their true dates. Editing any old workout re-replays history so records always match reality
- Atlas and the CLI now see your real record book โ actual lifted records with estimates explicitly labeled as estimates, plus the 1,000-lb club now sums your real big-3 singles instead of projections
Schedule Changes Never Rewrite History
v4.35.0- Changing a running schedule's dose or frequency no longer rewrites the past. Before: editing 200โ300 mg silently recalculated every previous week's blood levels as if you'd always been on 300. Now the old protocol is preserved (ended yesterday) and a new schedule starts today โ the serum curve keeps your real history and adjusts only going forward
- One continuous curve per compound: the old protocol's decaying levels and the new protocol's doses stack into a single line, exactly like your bloodstream does. Steady-state resets to ~4 half-lives after the change โ 'don't judge the new dose yet' now shows automatically
- Ending a compound no longer erases it from Active Levels โ the chart and dashboard now show its tail 'clearing' as it decays out of your system instead of vanishing instantly
- Works everywhere schedules are edited: the Schedules tab, Atlas dose adjustments, and the CLI all preserve history the same way. Editing an already-ended schedule still updates in place โ correcting the historical record stays possible
Custom Doses: Any Date, Real Peaks
v4.34.0- The custom dose dialog now uses a real calendar picker โ log a dose for ANY date, past or future, instead of being locked to a fixed 7-day window. Took an extra shot yesterday and forgot to log it? Pick yesterday and log it
- Off-schedule doses now count toward Active Levels: an extra 200 mg on top of your scheduled protocol shows up as a real peak in the serum-level curve, on the dashboard sparkline, and in Atlas's context โ your modeled blood concentration finally includes everything you actually took
- Custom doses for compounds you don't even have a schedule for get their own curve on the Active Levels chart, labeled as off-schedule doses
- Fixed the custom dose dialog overflowing its edges โ long compound names no longer push the dropdowns past the popup border
Sunday, Jul 26, 20262 releases
Atlas Stops Guessing Compound Facts
v4.33.0- Atlas no longer answers compound pharmacology (aromatization, chemical structure, mechanism, hepatotoxicity, ratios) from its own memory โ where it was confidently wrong (e.g. claiming Tbol 'retains residual aromatization' and inventing chemical structures). It now uses ONLY the app's compound library
- Every active compound now carries its full recorded pharmacology in Atlas's context โ aromatization/estrogenic activity, derivative class, anabolic:androgenic ratio, half-life, effects, side effects โ instead of a truncated blurb
- When a fact isn't in the library, Atlas now says so plainly instead of guessing โ and it will never fabricate a chemical formula or structure again
- New aromatization/estrogenic data recorded for testosterones, 19-nors, DHT-derivatives, SARMs, AIs, SERMs and more โ so questions like 'does Tbol aromatize?' get the correct answer (it doesn't โ the 4-chloro group blocks it)
- The lookup tool now returns a full fact sheet and Atlas is required to consult it before stating any compound property it doesn't already have
Atlas Reads Your Bloodwork Against the Right Doses
v4.32.0- Atlas now ties every recent lab panel to the exact protocol you were running WHEN THAT BLOOD WAS DRAWN โ no more assuming your current doses produced an older result. Your confirmed 'what I was on' for the latest panel is now in Atlas's context, not just the older-panel tool
- When your dose changed since a draw (e.g. Arimidex 0.25mg โ 0.5mg), Atlas is explicitly warned of the difference so it can't silently swap in today's dose when discussing that panel
- Confirmed panel protocols are treated as exact ground truth; unconfirmed ones are clearly flagged as estimates so Atlas hedges instead of asserting
- Fixed chat titles refusing: a stricter title model could turn 'What would raising my test do?' into a thread named 'I can't provide advice onโฆ'. Titles (and workout names) are now framed as labeling tasks, and any refusal is discarded in favor of your own words โ the coach itself was always answering
- Repaired the existing mis-titled conversation and hardened every place a model's raw output becomes visible text
Monday, Jul 20, 20261 release
Atlas Tells the Truth Now
v4.31.0- Atlas now sees your workouts set by set, exactly as performed โ no more compressing '8 @ 135, 4 @ 225, 8 @ 135' into a fictional '3ร4 @ 225'. Warmup/drop/failure sets and RPE are labeled too
- New built-in calculator: Atlas is now required to compute every number (dose totals, days of supply, volume, conversions) with an exact server-side calculator instead of doing mental math โ and to re-verify with it when you ask it to check
- Atlas now knows which model it runs on and reports it accurately when asked, instead of guessing
- Active compounds in Atlas's context now carry their description, effects, and mechanism inline โ better answers without a lookup round-trip
- Fixed a glitch where asking about older workouts could show a broken approval card in the chat
Sunday, Jul 19, 20266 releases
No More Unloggable Compounds
v4.30.0- Fixed the Weekly Log bug where compounds (and supplements) whose schedule end date had quietly passed showed as permanently greyed-out rows you couldn't log โ even though you were still taking them
- Ended or not-yet-started schedules in the Weekly Log now wear a clear 'Ended 6/15' / 'Starts 7/22' chip instead of an unexplained row of dashes
- New Resume button on ended schedules (Compounds & Supplements โ Schedules) clears the end date in one tap so you can keep logging
- Schedule end dates are now visible and editable โ set one, change one, or clear it to make the schedule open-ended
- Compound & supplement adherence no longer counts doses from ended schedules you physically couldn't log, and long-running schedules (2+ months old) count correctly again instead of silently dropping to zero
Your Streak Was Never Broken
v4.29.0- Weekend sessions now count: training Saturday or Sunday covers a skipped Friday, so swapping a Friday workout to the weekend no longer nukes your streak
- Workout dates are now computed in YOUR timezone โ late-evening sessions were being logged under the next day's date (UTC), which could silently break streaks and mislabel chart points
- Longest-streak and current-streak now use the same rules, so badges (like the 90-day Quarter Master) unlock off your real history
- Retroactive: streaks are computed from your full history, so any streak wrongly broken by these bugs is automatically restored
Latest Labs Card Actually Shows Your Latest Panel
v4.28.0- The dashboard Latest Labs card now shows your most recent panel โ every marker from one blood draw, stamped with its draw date โ instead of silently blending markers from different panels (which could surface an old E2 next to new results)
- Real navigation: back/forward chevrons, properly tappable page dots, and swipe support on mobile
- The moment you navigate manually, auto-rotation stops for good โ the card no longer cycles away from the page you picked
- Atlas's main context got the same fix: it now sees your full latest panel (up to 60 markers, was capped at 15) as one coherent snapshot, and the coach client overview leads with out-of-range markers
Delete a Whole Panel in One Click
v4.27.1- New trash button on every panel in History deletes the entire panel โ all markers plus its confirmed protocol โ in one action instead of 30 per-marker clicks
- Two-step confirm shows exactly how many markers will be removed before anything is deleted
Labs Are Dated by the Draw, Not the Upload
v4.27.0- Lab upload now reads the specimen collection date straight off your report and pre-fills it โ no more panels silently stamped with today's date when the blood was drawn weeks ago
- If the report doesn't state a collection date, you get a loud warning to set the draw date instead of a quiet default to today
- Panel dates are now editable after logging โ hit the pencil on any panel in History to move every marker (and its confirmed protocol) to the correct draw date
- The date field is now labeled "Blood Draw Date" so there's no ambiguity about which date it wants
Protocol Context Overhaul โ Tell Us Exactly What You Were On
v4.26.0- Every bloodwork panel's "what you were running" is now fully editable โ hit the pencil and record the exact compounds, doses, and frequencies you were actually taking when that blood was drawn
- Confirmed protocols are ground truth: they override the schedule-based guess everywhere, including what Atlas sees โ no more compounding inaccuracies
- Honest labels: unverified panels are clearly badged "Estimated โ verify" instead of being stated as fact, and Atlas now hedges on estimates instead of asserting them
- Confirming an empty list is supported โ explicitly mark a panel as drawn while off-cycle
- Changed your mind? Reset any confirmation to fall back to the schedule-derived estimate
Tuesday, Jun 30, 20262 releases
Active Levels โ Your Serum Levels, Now on the Dashboard
v4.25.0- New Active Levels dashboard card shows the estimated active amount of each injectable you're running right now, no need to open the full chart
- Each compound gets a mini sparkline โ solid through today, dashed for the projection โ plus its half-life and a live steady-state readout
- Compounds still ramping show how many days until they plateau; ones that have leveled off read "at steady state"
- Powered by the same first-order PK model as the Active Levels page โ educational, not medical advice
Smarter Labs Card โ Cycles Through Your Bloodwork, Tailored to Your Cycle
v4.24.0- The Latest Labs card now cycles through every biomarker from your most recent panel instead of showing a fixed five โ tap the dots or just watch it rotate
- Out-of-range markers always float to the top so nothing important gets buried
- Tailored to what you're running: on anabolics it leads with the hormonal panel (Testosterone, E2, SHBG, LH, FSH, hematocrit); on peptides/GLPs it leads with liver, lipids & metabolic markers
- Fixed the Recent PRs card showing your all-time strongest lifts instead of your most recent โ it's sorted by date again
Friday, Jun 26, 20261 release
Command Line Interface โ Let Your AI Agent Run Your Account
v4.23.0- Generate an API key in Settings โ Command Line Interface and hand it to Claude Code, a Telegram bot, or any AI agent โ it can now read and write your whole account
- 50+ tools: log workouts, injections, supplements, food, labs, weight & measurements; manage cycles, templates, schedules, goals; and look up any of your data on demand
- One dependency-free Python script (download it or grab it on GitHub) that an agent discovers and drives automatically โ everything Atlas can do, plus granular logging, from your terminal
- Keys are hashed, scoped to you, revocable anytime, and shown only once
Monday, Jun 22, 20261 release
Atlas Can Pull Up Any Past Workout
v4.22.0- Ask Atlas to compare to an older session โ 'how'd this stack up to my Monday last week?' โ and it now looks the workout up instead of saying it isn't logged
- Works for trends too: 'how has my bench moved this cycle' pulls that lift's history across any date range
- Fixes the coach previously only seeing your 5 most recent sessions, with no way to reach further back
Saturday, Jun 20, 20268 releases
Today's Macros on Your Dashboard
v4.21.0- New dashboard widget: today's logged calories and protein/carbs/fats against your targets, at a glance โ added automatically so it's there next time you open the app
- Rearrange or hide it like any widget in dashboard settings
Coaches See Atlas Flags
v4.20.0- The coach client view now shows Atlas's recent flags for that client โ bloodwork drifting, missed injections, off-target weight โ so coaches catch what the AI already noticed
- Warnings, nudges, and info are color-coded at a glance alongside the client's snapshot and labs
Plan a Cycle โ Model Before You Start
v4.19.0- New cycle planner (Compounds โ Active Levels โ Plan a cycle): build a hypothetical stack and instantly see the projected serum-level curves for the whole run
- Tune dose, frequency, and weeks per compound and watch the peaks/troughs and steady-state shift live โ before you commit a single pin
- Pure what-if: nothing is saved, so experiment freely. Educational model, not medical advice
Labs in Context โ See What You Were Running
v4.18.0- Your Labs page now shows a Protocol Context panel: for every bloodwork date, exactly which compounds you were running โ dose, frequency, and half-life
- โ It also flags recently-stopped compounds that were likely still in your system at draw time (within ~5 half-lives), so a marker change lines up with the real cause
- Same logic Atlas uses when it reads your lab history, now visible to you
Atlas Reads Your Serum Curve
v4.17.0- Atlas now turns your serum-level model into proactive coaching: if a compound swings hard between injections, it suggests dosing more frequently to flatten the curve
- And when you start a long-ester compound, Atlas tells you roughly when it'll reach steady state โ so you don't judge results too early
- These alerts flow into the same notification bell as the rest of Atlas's insights
Food Logging โ Close the Loop
v4.16.0- Log what you eat right on the Nutrition page: add items with calories and macros and watch your intake fill toward your targets with live progress bars
- Atlas now sees your intake, not just your results โ so coaching finally accounts for what's actually going in
- Simple and fast by design. Barcode scanning and camera-based macro estimates are on the roadmap next
Active Levels โ See Your Serum Curve
v4.15.0- New Active Levels view (Compounds โ Levels): a serum-level curve for every injectable you're running, modeled from your dosing schedule and each compound's half-life
- See the peaks-and-troughs sawtooth flatten toward steady state โ and exactly when each compound reaches it (โ4 half-lives in), so you know when results are worth judging
- Educational pharmacokinetic model (population-average half-lives, idealized absorption) โ not medical advice
Lab Copilot โ Atlas Watches Your Bloodwork
v4.14.0- Atlas now proactively flags bloodwork that's drifting the wrong way: a marker over its reference range, or climbing across your last three panels toward the top of range
- Watches the markers that matter most for enhanced athletes โ hematocrit, hemoglobin, ALT/AST, GGT, LDL, triglycerides, creatinine, estradiol, prolactin โ and surfaces the most concerning one
- Shows up as a warning in your notification bell (and a push if enabled), always pointing you back to your provider for clinical decisions
Thursday, Jun 18, 20263 releases
Coaching โ Bring Your Coach Into the App
v4.13.0- New Coach โ Client mode: share a coaching invite link, and once a client accepts, you can see their dashboard, body metrics, labs, and cycle in one place (Clients in the sidebar)
- Assign on their behalf: push macro targets, set a cycle goal weight, or copy one of your workout templates straight into a client's account
- Clients get an Atlas insight whenever their coach updates their plan, so nothing gets missed
- You're always in control of your data: coaching access is granted by accepting an invite and can be revoked anytime in Clients
Onboarding Fix โ Pasted Routines Import Fully
v4.12.2- Fixed setup dropping your workout templates when you pasted a long routine (or your whole conversation) in one message โ the full routine is now read end-to-end and every training day imports
Onboarding Fix โ Natty Setup No Longer Stalls
v4.12.1- Fixed a bug where finishing Atlas's setup interview could fail to save for users who don't track compounds โ the final step now completes reliably regardless of which sections you fill in
Wednesday, Jun 17, 20261 release
Atlas Reaches Out โ Proactive Coaching
v4.12.0- Atlas now notices things for you instead of waiting in the chat: weight drifting outside your cycle's target band, injections missed this week, or bloodwork that's overdue while you're on cycle โ all surfaced right in your notification bell
- Turn on push notifications (Settings โ Notifications) to get warnings and nudges on your device. Add the app to your iPhone Home Screen first to enable push there
- You're in control: mute any category, pick a weekly or daily digest, or switch proactive insights off entirely โ and Atlas speaks in your chosen coaching tone
- New weekly digest: a Monday summary of how your week went, right in the bell
Monday, Jun 15, 20261 release
Lean Mass Trend โ See the Recomp
v4.11.0- Lean Mass Trend now shows your recent change over the same ~7-day window as Weight Trend โ so next to a "+18 lb" weight gain you can see exactly how much of it was lean mass
- New recomp split breaks the period's total weight change into lean vs fat (e.g. +18.0 total โ +6.2 lean ยท +11.8 fat), doing the partition math for you โ and it instantly reveals a true recomp when lean goes up while fat goes down
- Hovering the lean-mass chart now shows weight and body-fat % for that day alongside lean mass, so every point tells the full story
Wednesday, Jun 3, 20263 releases
Voice That Works on iPhone
v4.10.2- Fixed the full voice conversation on iOS โ audio is now resampled to the rate Grok expects (iOS forced a different one), so Atlas can actually hear you on your phone
- "Speak to type" now uses fast server-side transcription (Groq Whisper) when available, so dictation works on iPhone/iPad where the browser's built-in speech engine simply isn't allowed
- Falls back to the browser speech engine automatically when server transcription isn't configured
Clearer Voice-to-Text Behavior
v4.10.1- The "speak to type" mic now explains itself when your browser blocks voice input (common in Firefox, Brave, and some installed-app modes) instead of failing silently โ with guidance to use Chrome/Safari or just type
- Longer grace period after tapping the mic so pausing to gather your thoughts no longer cuts the session off before you start talking
Coach Memory โ Atlas Remembers You Across Chats
v4.10.0- Atlas now keeps a persistent memory across every conversation โ tell it you tweaked your elbow or that you train fasted, and it remembers, dated, in future chats instead of forgetting the moment you start a new thread
- It saves notes on its own as you talk (injuries, symptoms, preferences, goals, life events) and weaves them back in proactively โ e.g. asking how an injury is healing, or factoring a known issue into why your volume dipped
- Transient notes auto-expire (a rough week fades on its own) while durable facts stick around; resolved issues get archived so they stop coming up
- A new "What Atlas Remembers" page (brain icon in the coach header) lets you review and delete anything it's saved โ you're always in control
- Built to stay light on context so long conversations keep running smoothly
Tuesday, Jun 2, 20265 releases
Twice-Weekly Schedules โ Pick Your Two Days
v4.9.8- The "Twice Weekly" frequency now lets you pick the exact two weekdays you dose on (e.g. Mon & Thu) instead of inferring them from your start date with uneven 3-then-4-day gaps
- Existing twice-weekly schedules keep working exactly as before โ edit one to choose explicit days whenever you like
Labs โ Cleaner History & Smarter Upload Progress
v4.9.7- Lab history is now organized into collapsible date cards instead of one endless list โ the most recent panel opens by default, and each date shows its marker count and how many are flagged out-of-range
- Uploading a report now shows a multi-step progress bar (uploading โ reading pages โ AI extraction โ double-checking) so you can see it's working during longer multi-page scans
Lab OCR โ Pinned to xAI's Best Extraction Model
v4.9.6- Lab report extraction now always uses Grok 4.3 โ the most accurate xAI model for this task โ no matter which model you've selected for the AI coach
- Chosen by a head-to-head bake-off on a 7-page, 44-marker panel: matched the newest reasoning model on completeness while running ~2.4ร faster
Grok/xAI โ Platform-Wide AI Reliability
v4.9.5- Standardized every AI feature on the xAI Grok endpoint โ AI coach chat, onboarding extraction, workout naming, exercise auto-fill, and lab OCR now all call the correct provider
- Retired the deprecated Grok 3 default everywhere โ stored model settings are auto-upgraded to the current Grok model at request time, so no account hits a 'model not found' error
- Fixed three AI endpoints that were silently routing xAI requests to the wrong provider (onboarding, exercise auto-fill, lab upload)
Lab Upload โ Reliable OCR for Large Multi-Page Reports
v4.9.4- Fixed lab report uploads failing on larger PDFs (7โ8 page cumulative panels) โ extraction now reliably reads 40+ markers in a single pass
- Lab OCR now calls the xAI Grok endpoint directly with the correct model, instead of misrouting to a provider that rejected the request
- Raised the analysis time limit so multi-page reports finish instead of timing out mid-way
- Clearer error messages that tell you what actually went wrong (timed out, too many results, unreadable file) instead of a generic failure
Sunday, May 31, 20262 releases
Date Picker โ Fix Clipping & Mobile Overflow
v4.9.3- Fixed the calendar popover being cut off vertically inside cards and dialogs โ it now renders in a top-level overlay that nothing can clip
- On mobile the calendar opens as a centered, full-width sheet with a backdrop instead of running off the edge of the screen
- The popover now flips above the field when there's not enough room below, and stays aligned while scrolling
Calendar Date Pickers Everywhere โ No More 'Invalid Date'
v4.9.2- Every date field across the app now uses the same friendly calendar picker โ compound & supplement schedules, weight & measurement logs, lab results, goals, photos, transformations, and data export
- Eliminated the browser's 'Please enter a valid date' error for good โ picking a date from the calendar always produces a valid value
- End-date fields are constrained to on/after the start date, so date ranges can't be entered backwards
- Consistent look, keyboard/Escape support, and a 'Today' shortcut on every date field